11.2% Efficient Solution Processed Kesterite Solar Cell with a Low Voltage Deficit

An 11.2% efficient kesterite solar cell is fabricated with a open circuit voltage (Voc ) deficit of only 0.57 V, which is possible by employing the three-stage annealing process. The reduced Voc deficit goes along with an increased minority carrier lifetime, low diode saturation current, and ideality factor, which are signatures of the semiconductor material with a low concentration of recombination centers. On the way towards a marketable and industrially-relevant photo­voltaic technology, kesterite – Cu2ZnSn(S,Se)4 (CZTSSe) – based thin film solar cells have to overcome several bottlenecks that currently limit their conversion efficiency. As of today, most of the synthesis routes for the CZTSSe absorbers have yielded solar cells with close to or above 10% power conversion efficiency. The current record of 12.6% has been obtained with a solution approach using hydrazine as the solvent,1 thermal co-evaporation of CZTSe yielded 11.6%,2 whereas devices fabricated by the selenization of sputtered metal precursors resulted in 10.4% efficiency.3 A comparison with the related chalcogenide Cu(In,Ga)Se2 (CIGS) technology reveals several deficiencies in CZTSSe solar cell parameters, while current research efforts are focused on the identification of bottlenecks that are limiting further improvements. Major limitations are the high open circuit voltage deficit (Voc-deficit) defined as the difference between the measured open circuit voltage and the absorber bandgap, and a comparably low fill factor (FF). Several possible explanations for the Voc-deficit have been proposed. Secondary phases such as binary and ternary copper-, zinc- and tin selenides and sulfides can exist in the absorber or its interfaces, which impede the carrier transport and lead to an increased recombination rate.4 Prevention of the formation of secondary phases in kesterite layers is challenging due to the small phase stable region for the formation of the quaternary CZTSSe compound.5 Losses in Voc can also be caused by recombination at the absorber interfaces due to a high density of surface states as well as cliff-like conduction band offsets.6 The lack of shallow defect states in CZTSSe is also expected to lower mobility, lifetime and therefore decrease the Voc.7 Finally, bandgap fluctuations stemming from structural or compositional inhomogeneities, or potential fluctuations due to a high concentration of charged defects can also lead to lower Voc.8 In this investigation, CZTSSe absorbers are prepared from precursors obtained from a solution approach with dimethyl sulfoxide (DMSO) as the solvent with subsequent recrystallization by annealing under controlled atmosphere of selenium. The DMSO approach,9 a safer and low-toxic alternative to the hydrazine route, requires only readily available metal salts and has yielded 10.3% efficient solar cells with a Voc deficit of 0.60 V10 or recently, even up to 11.8% measured on active area.11 Some commonly reported problems of the DMSO-processed kesterite layers are their high porosity, nonuniformity, and numerous grain boundaries that can lead to undesirable recombination.12 Here, we employ a three-stage annealing process under controlled selenium atmosphere in an SiOx coated graphite box to drastically improve the grain size and morphology of the absorber layer. Importantly, the Voc deficit can be reduced to 0.57 V, which appears to be one of the lowest values reported for kesterite devices. Systematic electrical characterization of absorbers and finished solar cells with time-resolved photoluminescence (TRPL), temperature-dependent current–voltage measurements (JV–T), and admittance spectroscopy (AS) are used to identify the reasons of the improved voltage. Figure 1 shows the scanning electron microscopy (SEM) cross sections of four different Cu2ZnSn(S,Se)4 (CZTSSe) absorbers A–D yielding efficiencies from 6.6% to 10.1% (total cell area of 0.3 cm2 including metal grid lines). The annealing conditions are varied from uncoated graphite box (sample A,B) to SiOx-coated graphite box (sample C,D) and two-stage temperature gradient (sample A,C) to three-stage temperature gradient (sample B,D); temperature gradients are presented in Figure S1 (Supporting Information). The selenization of sample A was conducted in an uncoated graphite box employing a two-stage temperature gradient, and the absorber layer exhibits a distinct bilayer structure with a thick small-grain bottom layer.13 Sample B was selenized in an uncoated graphite box similar to A but employing a three-stage temperature gradient. The SEM cross section shows an improved crystallization and grain size in both upper crust and bottom layer. However, the distinct bilayer structure of the absorber layer remains. The selenization of sample C was conducted in an SiOx-coated graphite box using the two-stage process, and the morphology of the film exhibits a comparably thin upper layer with small grains, but an improved crystallization in the bottom layer in contrast to sample A. Finally, sample D was selenized in the SiOx-coated graphite box with the three-stage temperature gradient and shows an overall improved crystallization with large grains and a significant reduction of the small-grain bottom layer. X-ray diffraction (XRD) pattern in Figure 2b shows a double kesterite reflex at 53.4° for all samples, indicating two regions with different S/(S + Se) ratio in the absorber layer. Grazing incidence XRD with varying incident angles confirms that the region with lower S/(S + Se) ratio coincides with the upper crust and the region with higher S/(S + Se) ratio belongs to the small-grain bottom layer. The reflexes corresponding to the higher S/(S + Se) ratio extenuate with the shift from uncoated to SiOx-coated graphite box as well as with the change of the temperature gradient from two-stage to three-stage process. No secondary phases can be identified from the XRD patterns of the four samples in Figure 2a, although the presence of Zn(S,Se) and Cu2Sn(S,Se)3 impurity phases cannot be excluded since their reflexes coincide with those of CZTSSe and therefore, cannot be distinguished by XRD.14 Reflexes at 14.7°, 17.4°, and 22.1° confirm unambiguously the CZTSSe phase. For all samples additional reflexes appear at 31.6° and 56.5°, indicating the Mo(S,Se)2 layer15 whose thickness depends on annealing conditions. Both changes in graphite box and temperature gradient increase the Mo(S,Se)2 layer thickness. The temperature gradient during the annealing step of the two-stage process features a lower holding temperature at 300 °C which is intended to outgas remaining residues of the organic solvent before the dense top crust on the absorber layer starts to form. Additionally, Na2Sex phases with low melting points can form, which act as fluxing agents and improve the grain growth.16 The higher holding temperature at 500 °C facilitates the crystallization, grain growth, and selenium incorporation. The selenium vapor pressure during the annealing equals the saturated vapor pressure for the given temperatures, that is, 0.3 and 55 mbar at 300 and 500 °C, respectively.17 The grain growth at 500 °C leading to the bilayer structure can be split into an abnormal grain growth of the large-grained upper layer and a normal grain growth for the small-grained bottom layer.18 The abnormal grain growth is driven by the surface area and total energy difference, and it is also facilitated by volatile Zn-, Sn-, and Se-containing species from the gas phase. The third holding temperature at 550 °C exclusively at the end of the three-stage process accounts for increased grain growth, however at this stage no additional selenium is left in the graphite box and the grain growth is presumably governed by redistribution of grain boundaries. The intention behind the SiOx coating is to significantly reduce porosity of the graphite box, thus enabling a higher selenium pressure during annealing, which improves crystallization especially in the bottom layer of the absorber. It should be noted, however, that higher temperatures and/or longer annealing times promote the formation of thicker Mo(S,Se)2 layers, that at some point can cause delamination and decrease device performance. The change in metal ratios due to the high-temperature annealing is proportional to the annealing time and corresponding temperature. The loss of Sn and Zn is more pronounced for the three-stage process compared to the two-stage process, due to the formation of volatile phases of both constituents.19 Surprisingly, the change from uncoated to coated graphite box did not affect the final metal ratios. As a result, metal ratios of samples A, C and B, D are quite close, Cu/(Zn + Sn) = 0.90 ± 0.03 and Zn/Sn = 1.27 ± 0.04, which are in the compositional range typically yielding the high efficiencies.20 In this off-stoichiometric range, so-called B-type, the presence of defect complexes [2Zn2+Cu + Zn2+Sn] with a formation energy of 0.86 eV is expected.21 Figure 3a shows the J–V characteristics of devices A–D and their basic photovoltaic properties. Changing the reactor from uncoated to SiOx-coated graphite box leads to a higher fill factor (FF), whereas the change from the two-stage to the three-stage temperature gradient during selenization improves FF, Jsc, and significantly decreases the Voc deficit. In Figure 3b, the external quantum efficiency (EQE) shows the spectral composition of the Jsc together with bandgap values Eg extracted from the inflection point of the EQE spectrum in the long-wavelength range. Integration of EQE spectra yields photocurrent of 31.6, 32.8, 28.2, and 35.4 mA cm−2 for samples A–D, respectively, which is in good agreement with Jsc values obtained from J–V measurements. The extracted bandgaps are similar to bandgaps estimated from XRD measurements assuming the bandgap of 1.0 eV for CZTSe and 1.5 eV for CZTS. Figure 4 shows the TRPL decay curves and the PL spectra acquired for devices A–D at room temperature. The TRPL decay curves indicate an increased radiative lifetime for the samples annealed with the three-stage process. The radiative lifetime correlates with the thickness of the large-grain upper layer of the absorber. The TRPL results are in good agreement with the long-wavelength characteristics of the corresponding EQE spectra, so that a lower EQE is observed for the samples with a faster TRPL intensity decay. The PL spectra exhibit a broad peak for all samples, and the peak maxima are redshifted as compared to the bandgap by 0.11, 0.02, 0.06, and 0.02 eV for samples A, B, C, and D, respectively. In literature, PL maxima redshifts of 110 meV are reported for solution-based methods, whereas vacuum methods can yield values as low as 10–20 meV.2, 3, 22 The rather broad shape in our case indicates the presence of tail states and/or potential fluctuations,8 which are responsible for the increased lifetime by several orders when performing low-temperature TRPL measurements (not shown here). The decrease in redshift of the PL maxima with respect to the bandgap values of samples A, C to B and D suggests a reduction in tail states. Figure 5 shows the properties of the champion device D_2, which was fabricated with an identical process as sample D but in another batch. The SEM cross-section image exhibits a similar structure as sample D with a large-grain upper crust, a rather narrow small-grain bottom layer, and a distinct MoSe2 layer. The JV–T curve yields a total area efficiency of 11.2%, whereas ten best cells had an average efficiency of 10.6 ± 0.3%. The EQE measurement shows that the major part of additional current is stemming from the long-wavelength region, which is in line with the long TRPL decay in Figure 5d. The decay curve is fitted with a single exponential function in the range 10–100 ns since the faster decay during first 10 ns originates from the charge separation in the device.22 The fitted long minority carrier lifetime of τ2 = 8.1 ns is responsible for the improved collection in the long-wavelength region, which is manifested by plotting the ratio of reversed bias EQE and zero bias EQE. Further advanced characterization on sample D_2 was conducted using AS and JV–T measurements. Figure 6a shows the JV–T curves in darkness and under illumination. The crossover of illuminated and dark curves is becoming more pronounced at lower temperatures, whereas the increasing rollover of the J–V curves leads to a complete blocking of the current at the lowest temperature of 123 K. Possible explanations for this blocking is a barrier at the interface between absorber and the Mo back contact, which facilitates the minority carrier recombination,23 or an increase in bulk resistivity, due to the lack of shallow acceptor states and therefore a freeze out of deeper acceptor states rendering the device fully depleted and exhibiting high resistivity.7 The temperature dependence of Voc extrapolated to T = 0 K provides an intercept of EA (Voc − T) = 0.99 eV, representing the activation energy for the dominant recombination mechanism. Since this value is very close to the derived bandgap of 1.05 eV, one can conclude that the dominant recombination paths are located within the bulk of the absorber rather than at the interface.24 However, the freeze out observed in the AS measurement could also be attributed to a Schottky barrier in the device and a corresponding activation energy is derived by a change in the temperature dependence of the prefactor from T2 to T3/2, yielding an activation energy of EA,B (Cf−T) = 115 meV.26 From the Cf–T measurements the dark series resistance can also be calculated by employing an admittance circuit model of a depletion region in series with the undepleted quasi-neutral region.7 Comparison of these values with the Rs derived from JV–T (Figure S2, Supporting Information) shows one order of magnitude lower values for Rs obtained from AS, implying that the activation energies EA (Cf−T) and EA (Rs−T) are not attributed to the same effect. In Table 1, published champion devices having efficiency >10% for nonvacuum and vacuum deposition techniques are displayed. The record efficiency of 12.6% is obtained with a pure-solution approach utilizing hydrazine as the solvent.1 The highest FF with 69.8% is also observed for this sample. DMSO-processed cells reach efficiencies of 11.2% (this study), 11.8% (active area measurement),11 and 10.3%.10 Vacuum-based pure selenide absorbers can deliver the highest photocurrents Jsc of 40.62 and 39.7 mA cm−23 because of their low bandgaps. The 10.8% mini-module from Solar Frontier features the lowest Voc deficit of 0.56 V (determined from a linear fit of the transformed EQE curve).27 The 11.2% device of this study exhibits the highest minority carrier lifetime of τ = 8.1 ns indicating a low concentration of defects in the absorber bulk and interfaces and the lowest ideality factor of A = 1.14 suggesting that the dominant recombination paths are shifting from the space charge region (SCR) to the quasi-neutral region (QNR),23 which is consistent with the fact that the Voc–T fit intersects at T = 0 K very close to the bandgap value. In summary, an 11.2% efficient (total cell area measurement) CZTSSe solar cell is fabricated using the hydrazine-free DMSO solution approach. The best device features the Voc deficit of only 0.57 V which is among the lowest for solution-processed CZTSSe devices. The open circuit voltage improvement was possible because of the three-stage annealing process in a silica-coated closed reactor, which enabled an increased incorporation of selenium and large-grained microstructure. The reduced Voc deficit is attributed to an increased minority carrier lifetime, low diode saturation current, and ideality factor, which are signatures of the semiconductor material with a low concentration of recombination centers. The CZTSSe precursor solution consisted of thiourea (99%+, Sigma-Aldrich), SnCl2·2H2O (98%, Sigma-Aldrich), ZnCl2 (99.99%, Alfa Aesar), CuCl2 (98%+, Alfa Aesar), and NaCl (99.99%, Alfa Aesar), dissolved in DMSO (99.9%, Alfa Aesar). The precursor solution contained 0.56 M CuCl2, 0.41 M SnCl2·2H2O, 0.44 M ZnCl2, 1.85 M thiourea and 0.15 M NaCl. The precursor solution was spin-coated onto an Mo-coated soda lime glass and dried on a hotplate at 320 °C in air. The spin-coating and drying steps were repeated 14 times in order to obtain the desired precursor film thickness of 1.5–2 μm. The precursor composition for all samples was Cu/Zn = 1.48, Cu/Sn = 1.62, and Zn/Sn = 1.10. Samples A and B were annealed in an Annealsys AS ONE 150 RTP furnace inside a closed graphite box with selenium pellets (800 mg), and samples C and D were annealed in the same furnace but inside an SiOx-coated graphite box with the equal amount of Se. The temperature gradient employed for annealing samples A and C was the two-stage process with holding at 300 and 500 °C whereas samples B and D were annealed with the three-stage process with holding at 300, 500, and 550 °C. Metal ratios of the selenized absorbers were for samples A and C: Cu/Zn = 1.57, Cu/Sn = 1.96, Zn/Sn = 1.25, for samples B and D: Cu/Zn = 1.62, Cu/Sn = 2.11, Zn/Sn = 1.30, and for sample D_2: Cu/Zn = 1.52, Cu/Sn = 1.92, and Zn/Sn = 1.26. All metal ratios were obtained by X-ray fluorescence (XRF) measurements calibrated by inductively coupled plasma mass spectrometry (ICP-MS). The temperature–time profiles for all samples are given in Figure S1 (Supporting Information). After selenization the absorbers were immersed for 30 s in a 10 wt% KCN solution in order to remove copper-rich phases and to clean the surface from contaminations and oxides. A 50–70 nm thick CdS buffer layer was deposited by chemical bath deposition, and 70 nm/350 nm i-ZnO/Al:ZnO bilayer was sputtered. An Ni/Al top grid and an AR coating of MgF2 were deposited by e-beam evaporation. Individual solar cells were mechanically scribed to an area of 0.30 ± 0.02 cm2. Finally, post-annealing of complete devices was performed at 120 °C for 60 min in air. SEM images were taken on a Hitachi S-4800 electron microscope, XRD patterns were recorded in 2θ\θ scan mode using a Bruker D8 diffractometer with CuKα radiation (λ = 1.5418 Å, beam voltage: 40 kV, beam current: 40 mA, calibrated using Si100 and Si111 single crystals), a step size of 0.04°, and a scan rate of 0.5 s step−1. The J–V characterization was performed under standard test conditions (100 mW cm−2, 25 °C, AM1.5G) using a solar simulator calibrated with a certified Si diode. The EQE spectra were recorded using a chopped white light source (900 W halogen lamp) with a LOT MSH-300 monochromator, which was calibrated with certified Si and Ge diodes. The illuminated area on the sample was 0.1 cm2 including grid lines. AS measurements were carried out with an LCR meter from Agilent (E4990A) with an AC voltage of 50 mV in the temperature range from 123 to 323 K. Room temperature TRPL and PL spectra were measured on an FT300 fluorescence lifetime spectrometer from PicoQuant with a 639 nm pulsed diode laser as excitation source (pulse width 90 ps, repetition rate 10 MHz) and a thermoelectric cooled Hamamatsu NIR-PMT module H10330A-45 (rise time 0.9 ns, transit time spread 0.4 ns). This research was supported by the 7th Framework Programme under the project KESTCELLS (FP7-PEOPLE-2012-ITN-316488). The authors would also like to thank the whole team of the Laboratory for Thin Films and Photovoltaics. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.